زبان و تجارت خارجی

While language plays an important role in gravity models, there has been little attention to the channels through which a common language promotes bilateral trade. This work proposes separate series for a common language depending upon whether ease of communication facilitates trade through translation or the ability to communicate directly. The series related to direct communication is far more important in explaining bilateral trade, but the other series, based on translation, makes a distinct contribution as well. Either measure of a common language outperforms the measure in popular use, which is implicitly related to translation, and a combination of the two does far better. In addition, the paper examines the effect of two country-specific linguistic influences on trade: Literacy and linguistic diversity at home. Both of these influences promote foreign relative to domestic trade. Finally, the article studies the separate roles of English and network externalities.

Gravity models provide ample evidence that a common language has a significant impact on bilateral trade but studies of these models leave open the question of the channels through which the effect takes place.1 The typical step is to use a binary measure of a common language as zero or one. But the channel of influence that this concerns is not obvious. Direct communication (DC) cannot be the entire answer since the measure often treats a common language as present when only a minority in either country in a pair could communicate directly with the people in the other. For example, India and Tanzania are supposed to be an English pair and Niger and Senegal a French one. But if we consider the probability that a random pair of people from India and Tanzania both speak English or a similar random pair from Niger and Senegal both speak French, the figure is less than 10 percent in both cases. Implicitly, a widespread ability to communicate directly is not essential and an adequate system of translation will do as well. In this paper, I propose to construct a separate series for a common language that depends on translation, and a separate series that depends on DC. This will make it possible to see whether translation acts a separate channel, as distinct from DC, through which ease of communication promotes bilateral trade. Better measures of a common language will result as well. The usual reliance on binary variables as indicators of a common language probably reflects mostly the difficulty of quantifying the numbers of speakers of different languages in a country. Yet considerable headway is possible by relying on Grimes (2000), now in its 14th edition since it first appeared in 1951. This work is the result of a massive effort to condense the information supplied by the entire profession of ethnologists about world languages. There have indeed been at least three recent efforts to use this work to construct a general quantitative index of language in economic research: Hall and Jones (1999), Wagner (2000) and Rauch and Trindade (2002). But the aim, in these cases, has never been to provide a general index of the ability to communicate directly in foreign trade. Hall and Jones focus on language out of a concern with certain institutional/legal features. In close connection, they limit their attention to a few major languages. Though concerned with communication, Wagner deals strictly with the trade of Canadian provinces, and chooses his languages accordingly. Rauch and Trindade focus on ethnic ties. Consequently, they collect data strictly on common native languages, whereas, of course, bilingualism is of the essence in regard to communication. There has been one earlier effort to construct a general quantitative index of a common language in order to analyze world trade from a similar perspective as mine, by Boisso and Ferrantino (1997). But the authors proceed, in this pioneering work, like Rauch and Trindade, to attribute only a single language to each person (not necessarily the mother tongue). They also rely on a far more summary treatment than Grimes’, by Katzner (1986). Besides the issue of the channels of influence of a common language, I propose to deal with a number of other major questions about the influence of language on foreign trade. One is the question whether the world's dominant language, English, is more effective than the rest in promoting trade. Another is the impact of linguistic diversity at home and literacy on foreign trade. As I will argue below, in principle, both of these country-specific aspects of language should promote foreign trade. I will also consider the issue of the possible network externalities of a common language. The basic results can be summarized as follows. First, a common language promotes international trade via translation as well as through DC. DC is far more important. But, very significantly, an established network facilitating access to a language plays a basic role too. These considerations lead to a higher estimate of the influence of a common language on foreign trade than found in previous work. Next, despite the dominant position of English as a world language, English is no more effective in promoting trade than other major European languages. On the other hand, the major European languages as a group (including English) are more efficient than other languages in promoting trade. Further, both literacy and a diversity of tongues at home do indeed boost foreign trade, in accordance with theory. Finally, I find no evidence of network externalities of language. However, the search for such evidence yields one important result, namely, that a scarcity of advantages of a common language leads to more intensive exploitation of those advantages of a common language that do exist. The discussion will begin by examining the construction of the linguistic variables and the theoretical basis for their influence. Section 2 will develop the gravity equation that will serve in the empirical analysis. It will also deal with the problem of estimation and the choice of estimation procedure. Section 3 will then present the econometric results for the influences of the bilateral variables in the model, including DC and indirect communication via translation. Section 4 will next proceed to introduce the separate country-specific influences in the model. Following, Section 5 will explore the network externalities of a common language. The final section will add some general discussion and suggestions for future research. All the raw linguistic series, including those I constructed, follow in a separate data Appendix (Table A1).

We knew beforehand that a common language promotes trade. This study sheds light on the channels through which language exerts this influence and introduces a number of effects of language that had gone unnoticed before, or at least had not been studied. Apart from DC, translation is important in surmounting linguistic obstacles. Though few people may speak Portuguese in a country, they may be “tuned” to the language all the same. Thus, the impact of a common language on foreign trade can issue from translation as well as the ability to communicate directly. Notwithstanding, DC appears about three times more effective than indirect communication in promoting trade. Taking both direct and indirect communication into account and measuring them with more care than usual, we find an impact of a common language nearly twice as high as before. Two general linguistic influences on foreign trade also emerge in the study. If people can read and write in any language, they can cope better with the problems posed by foreign languages in general. In addition, if they face linguistic obstacles at home, they trade more with foreigners. Accordingly, based on identical scores for common language (notably including zero), the pull of linguistic factors in favor of domestic trade is uneven and is lower for country pairs sharing high literacy rates or sharing high linguistic diversity or both. Further, English seems to offer no particular advantage in foreign trade, while the European languages as a whole do. These languages emerge as better instruments of OCC than other ones. In addition, there is no evidence that network externalities of language contribute to trade between countries. The importance of those externalities has thus been exaggerated. The reason may have to do with the fact that a lot of market information gets transmitted without a common language (through limited vocabularies, small numbers of fluent speakers, dictionaries, and means of non-verbal communication). The separate significance of literacy and linguistic diversity would also argue that communication does not break down entirely without a common language.22 The study suggests a number of avenues of further research. First, it would be interesting to see if the relative significance of OCC and DC differs for trade in different sorts of goods. As a possibility, rudimentary communication might suffice for trade in homogeneous goods while trade in heterogeneous goods might require more sophisticated intercourse. If so, OCC might be particularly important in trade in homogenous goods while DC is unusually so in trade in heterogeneous goods. Rauch (1999) suggests a hypothesis of this sort. He argues that close personal relations between exporters and importers are much more significant in trade in heterogeneous goods than perfectly homogeneous ones. Because of issues of information, private “networks” are necessary in one case whereas “markets” suffice in the other. The proposed hypothesis about the separate roles of OCC and DC in different markets follows easily. and Another basic extension of the study would be to admit additional variables reflecting culture or ethnicity. Language has been interpreted here strictly as a tool of communication, even though it obviously reflects many aspects of culture as well. The reason for this narrow interpretation is that, in the context of worldwide trade, the only features of the language variables that apply generally regard communication. Other associated features are not always present, as it would be easy to document. This is obvious in the case of OCC where translation is the central mechanism. But even as regards DC, I have taken steps to assure the primacy of communication by including second-language and non-native speakers. Based on other work in the literature, one missing cultural variable that easily comes to mind is the stock of immigrants. As the analysis here stands, immigrants are implicit since they affect the index of linguistic diversity. But this index also reflects longstanding, sometimes ancestral, linguistic divisions inside national boundaries (as in the case of India or Switzerland, for example). The interest of a separate consideration of immigrants would lie in setting apart the element of ethnic ties to other communities abroad (which affects tastes, skills, trust and information) from strict issues of communication. A separate treatment of immigrants would generally promote the interpretation of the linguistic variables as relating exclusively to communication (compare note 24). A further extension of the study would be to examine trade in goods that are especially connected to language, such as movies, television programs, books, and vocal music. In these cases, the world dominance of English and the production by the US in particular, is notorious. Cultural products in English sell extremely well in many places where English is otherwise secondary in foreign as well as domestic trade. Broadly, how come English plays no special role in facilitating foreign trade but dominates the market for language-related products?25 Evidently, the question cannot be studied within the confines of the current version of the gravity model, which applies strictly to two-way trade, but would require a more intricate version of the model that distinguishes between exports and imports (and is often associated with Bergstrand (1985)).26 Finally, it would be interesting also to move beyond the idea that the language variables are constants. Literacy rates have risen substantially in the last 50 years. Migrations accelerated in the nineties in many parts of the world. Spanish is now significant in the US, Russian in Israel, etc. In treating the data furnished in Grimes (2000) as roughly contemporaneous and steady, I have followed the practice in the field of trade of viewing all indicators of a common language and linguistic diversity in foreign trade as slow-moving variables that can be regarded as fixed.27Grimes (2000) also corroborates this practice by furnishing unique figures for linguistic variables despite wide discrepancies in dating when dates even appear. Yet from the standpoint of trade analysis, there is little doubt that more coherent series for literacy, OCC and DC could be constructed from a broad variety of national sources. Doing so would foster study of the evolution of all three variables and their influences on trade.